Bytedance’s AI Push Stalls as Compute Limits and Copyright Issues Bite

ByteDance’s rollout of Seedance 2.0 has already run into a severe compute bottleneck that is throttling the model’s promised speed. According to Wired, users who manage to log into one of ByteDance’s domestic AI apps are often placed in queues of tens of thousands; a test this week placed a requester at position 90,985 with an estimated four‑hour wait for a five‑second clip, and even after two hours the video remained unrendered. The company has not disclosed the size of its GPU farm, but the queue length suggests that the infrastructure cannot keep pace with demand, a problem that is likely to worsen as more developers await API access. IT Home, citing ByteDance’s newly posted pricing sheet, notes that the current limit is a 15‑second video at a cost of just over $2, underscoring that the service is still constrained to short outputs while the hardware crunch persists.

The compute squeeze is compounded by regulatory headwinds that have limited ByteDance’s access to the most powerful AI accelerators. Reuters reports that Chinese regulators have barred the firm from using Nvidia chips, a restriction that directly curtails the company’s ability to scale its video generation workloads. Without Nvidia’s high‑throughput GPUs, ByteDance must rely on domestically produced alternatives that, according to industry analysts, lag behind in performance for large‑scale generative models. The combination of a capped hardware supply chain and an exploding user base creates a classic supply‑demand mismatch, forcing the company to ration compute time and keep the model’s output length at a modest 15 seconds.

Legal challenges are emerging in parallel with the technical hurdles. Wired notes that major Hollywood studios—including Disney, Netflix and Paramount—have issued cease‑and‑desist letters alleging that Seedance 2.0’s generated footage infringes on copyrighted works. The letters claim that the model, trained on vast swaths of publicly available video, reproduces recognizable scenes, characters and visual styles that belong to the studios, raising immediate liability concerns for ByteDance. ByteDance has not responded to requests for comment, but the legal pressure could force the firm to impose stricter content filters or limit public access, further dampening the model’s commercial rollout.

Despite the obstacles, ByteDance appears to be inching toward broader availability. Wired reports that the company updated its API platform this week and disclosed tentative pricing, suggesting that third‑party developer access is on the near‑term horizon. Afra Wang, who tracks the US‑China AI divide, points out that while China lags behind the United States in AI coding tools, it “is miles ahead” in video AI, and Seedance 2.0 could cement that lead if the compute and legal issues are resolved. However, the current queue times and the looming copyright disputes mean that the model’s impact will be limited to a small, privileged cohort of domestic users and resellers who are willing to sell their accounts abroad, as noted by Wired.

The broader strategic implications are stark. ByteDance’s ambition to dominate AI‑generated video hinges on scaling Seedance 2.0 beyond its current 15‑second, $2‑per‑clip niche while navigating a hostile regulatory environment and an aggressive legal offensive from Western media giants. If the compute bottleneck persists and the copyright lawsuits force stricter safeguards, the model may never achieve the mass‑market penetration needed to justify the investment. Conversely, a successful resolution—through domestic chip advancements, expanded GPU capacity, or a licensing framework with content owners—could give ByteDance a decisive advantage in a market where Western competitors such as OpenAI’s Sora remain nascent. For now, the promise of “thinking like a director,” as praised by Chinese creator Pan Tianhong, remains tempered by the very real constraints of hardware, policy and intellectual‑property law.